Anthropic builds its own chips and hires a Supreme Court justice as CGAO while OpenAI opens ChatGPT Health to every US adult: both labs are scaling in very different directions.
Anthropic builds its own chip team and hires a Supreme Court justice
Anthropic confirmed the launch of an in-house chip design team, a significant move toward building hardware tailored to its own models rather than depending entirely on Nvidia. In the same week, Mariano-Florentino Cuéllar, a former California Supreme Court justice and president of the Carnegie Endowment for International Peace, joined Anthropic as Chief Global Affairs Officer. Both announcements arrived ahead of Anthropic's anticipated IPO.
Why it matters: When an AI lab starts designing its own chips and hiring senior diplomats, it is not just building a product, it is building an institution. Anthropic is signalling that it expects to operate at a scale and political complexity that requires both, for the long term.
OpenAI rolls out ChatGPT Health to every US adult, for free
One day after a lawsuit was filed challenging it, OpenAI pushed ahead and made ChatGPT Health available to all US adults at no cost. The feature offers AI-guided health information and symptom context to anyone with a free account. Separately, OpenAI also announced that text chat will become fully unlimited for free users, removing one of the last remaining restrictions on the free tier.
Why it matters: AI health access is now a free public utility in the United States. That is a meaningful shift in who can get basic health information, and it will intensify regulatory scrutiny around AI medical advice in India and elsewhere.
Google restructures its AI division for the first time in a decade
Google announced a fundamental reorganisation of its AI division, the first major structural change to an organisation that has been running in roughly the same form for ten years. Details of the new structure were not fully disclosed, but the move is widely seen as aligning Google's AI research and product teams more directly under Gemini's commercial roadmap.
Why it matters: A decade-long structure getting torn down is not routine housekeeping. It suggests Google believes its previous model of separating research from product is no longer working at the pace the market demands. The companies that move fastest on this alignment will likely ship better products.
National Grid pays $1.75B for a stake in an AI power company
National Grid invested $1.75 billion for a 35% stake in Joulent, a US energy company whose core business is supplying reliable electricity to AI data centres. The deal reflects growing pressure from hyperscalers who are struggling to secure enough clean, dependable power to run the next generation of AI infrastructure at the scale they need.
Why it matters: The energy constraint on AI is now attracting billion-dollar investment from traditional infrastructure players. If you work in energy, sustainability, or climate, AI is about to become one of the biggest pressures on your sector, and the biggest opportunity.
India's MeitY leads a roundtable on how to fund AI startups
Cabinet minister Jitin Prasada led a roundtable under the IndiaAI Mission bringing together venture capital firms, global tech companies, and AI startups to address the funding gap for early-stage Indian AI builders. The session focused on compute credits, application grants, and a Fund of Funds structure designed to back domestic AI companies. The IndiaAI Mission's total outlay is ₹10,371 crore across seven pillars covering compute, datasets, research, and skilling.
Why it matters: India now has official government infrastructure for funding AI startups, not just policy papers. If you are an early-stage builder or researcher in India, the window to access compute credits and grants through this programme is open right now.
Qwen3.8-Max ships as a 2.4-trillion-parameter open model
Alibaba's Qwen team released Qwen3.8-Max, a 2.4-trillion-parameter mixture-of-experts model with 95 billion active parameters, a 1-million-token context window, and native video input support. The model is available openly and priced at $0.03 per million input tokens on the base tier, putting frontier-class capabilities within reach of developers who cannot afford closed-model API costs.
Why it matters: A 2.4-trillion-parameter open model with a 1M context window and video input changes what is possible for developers working outside the big cloud providers. The open-source tier of AI is no longer a step behind the frontier, it is arriving at the frontier itself.
Anthropic is preparing to become a public company, and everything shows it
Claude Opus 5, an in-house chip team, a former Supreme Court justice as Chief Global Affairs Officer, and an anticipated IPO on the horizon. Each of these moves makes sense individually. Together they describe a company that is done being a research lab and is actively building the kind of institutional credibility, infrastructure independence, and policy relationships that a listed company needs. Watch how quickly Anthropic's positioning shifts from AI safety pioneer to regulated financial entity.
OpenAI quietly introduced GPT-5.5, featuring a native "System 2" architecture, a design that separates fast, automatic responses from slower, deliberate reasoning within a single model. Separately, fallout continued from OpenAI's recent AI cyber capability tests, with questions being raised about disclosure timelines and who was notified and when.